Reduce typing latency caused by POLL_WAIT

[yizhepku]

This PR makes it so that Reedline no longer wait for POLL_WAIT after every keystroke. Instead, the delay is only triggered if text pasting or resizing is detected.

It also re-introduces a fix from PR #651. That PR was reverted by #840 because of issue #757 about external printers. If the feature external_printer is not enabled, this PR will just block on input forever for maximum power efficiency. Users with external_printer enabled will also see a small improvment in power efficiency, because this PR makes it possible to use a longer polling inteval without affacting typing latency.

Fixes #643, #10484, and #521, and hopeful doesn't break #757.

[yizhepku]

PR draft #737 discussed reworking the external printer API. I wonder how is that going?

[sholderbach]

Nushell doesn't make use of the external printer, so politically it takes precedence in that decision in my view. We should prioritize the battery saving behavior and figure out a solution for a responsive external printer in due course. (as the stalling of #737 shows, probably non trivial refactor needed.)

[yizhepku]

We should prioritize the battery saving behavior and figure out a solution for a responsive external printer in due course.

This PR aims to be maximally power efficient when the feature external_printer is NOT enabled. External printers are still responsive when enabled. It's the best of both worlds!

[fdncred]

Thanks. Let's move forward with this since it's early in our release cycle.